#!/usr/bin/env python
import numpy as np
# my custom function for generating a dataset
from nedc_ecg_gen_feats import ret_tens
from nedc_ecg_models_functs import define_resnet,define_base
from keras.utils import to_categorical

# Global variables
RESOLUTION = (1280,960)

# testing bottleneck
import time

def main():

    
    # define model parameters
    modeltype = "base"
    epochhs = 20
    learnrate = .01
    model = None
    numfilters=50
    filtersize=(20,8)

    tensors,labels = ret_tens()
    print(type(labels))
    print(set(labels))
    classnum = len(list(set(labels)))
    # defin the model
    model = define_base(64)
    modelname = "models/E20LR0005"
    #print("Model out = ",modelname+".keras")
    
    # fit and save the model

    start = time.time()
   # print (len(labels))
    print(to_categorical(labels,num_classes=64))
    history = model.fit(np.array(tensors),to_categorical(np.array(labels),num_classes = 64),epochs=epochhs)
    
    end=time.time()
    model.save(modelname+".keras")
    #print("model outputted")
    
    print("Time to process",len(labels),"files = ",end-start,"seconds")

if __name__ == "__main__":
    main()